Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 REVIEWS We discuss the development of dynamic 3D bioreactor-based systems as in vitro models for use in DMPK studies. Reviews  FOUNDATION REVIEW The use of bioreactors as in vitro models in pharmaceutical research Maaria Ginai1, Robert Elsby2, Christopher J. Hewitt1, Dominic Surry2, Katherine Fenner2 and Karen Coopman1 1 Centre for Biological Engineering, Department of Chemical Engineering, Loughborough University, Loughborough LE11 3TU, UK 2 In vitro In Silico ADME, Global DMPK, AstraZeneca Research and Development, Alderley Park, Mereside, Macclesfield SK10 4TG, UK Bringing a new drug to market is costly in terms of capital and time investments, and any development issues encountered during late-stage clinical trials can often be the result of in vitro–in vivo extrapolations (IVIVE) not accurately reflecting clinical outcome. In the discipline of drug metabolism and pharmacokinetics (DMPK), current in vitro cellular methods do not provide the 3D structure and function of organs found in vivo; therefore, new dynamic methods need to be established to aid improvement of IVIVE. In this review, we highlight the importance of model progression into dynamic systems for use within drug development, focusing on devices developed currently in the areas of the liver and blood– brain barrier (BBB), and the potential to develop models for other organ systems, such as the kidney. Currently, the pharmaceutical industry balances the stringent testing of drugs and products against soaring costs and return on investments risks. Bringing a new drug to market will, presently, on average cost US$1.3 billion over 12 years [1,2] with preclinical in vitro testing consuming approximately half of the total development time [2]. Moreover, many new chemical and biological entities (NCEs and NBEs, respectively) that fail late-stage human testing provide evidence for the fact that pharmacological and toxicity data from in vitro cell-based assays are not always predictive of the clinical situation [3]. Owing to these financial and time commitments, there is great emphasis within the industry on the development of newer and more reliable in vitro and preclinical methods to accompany or replace the existing methods of NCE–NBE investigations. In vitro models, particularly cell-based versions, are used in various areas of drug discovery, such as target identification and validation using disease models, compound screening and basic cytotoxicity using static cultures, through to absorption, distribution, metabolism, excretion and toxicology (ADMET) studies on lead compounds. Although high-throughput cell-based assays have revolutionised the efficiency and speed at which compounds can be screened, ADMET models ideally should be an accurate representation of the physiological or pathophysiological Maaria Ginai is currently studying a BBSRC CASEfunded PhD at the Centre for Biological Engineering at Loughborough University (UK) in the area of bioartificial device progression to in vitro models for use in the pharmaceutical industry. Her PhD is supported by AstraZeneca. She graduated with a first-class Bachelors degree with honours in biomedical science from the University of Kent (UK) in 2010. Christopher J. Hewitt is the director of the £7.3 M EPSRC Doctoral Training Centre in Regenerative Medicine and co-founder of the £2 M Centre for Biological Engineering at Loughborough University (UK). He also leads the Cell Technologies research group, whose work spans the engineering–life science interface seeking to understand the interaction of the organism with the engineering environment within such diverse areas as microbial fermentation, biotransformation, cell culture and, mostly recently, regenerative medicine bioprocessing. Christopher has a first-Class Bachelors degree in biology from Royal Holloway College, University of London (UK) and a PhD in chemical engineering from the University of Birmingham (UK). Karen Coopman was appointed to a lectureship at Loughborough University, where she is the operations manager of the EPSRCfunded Doctoral Training Centre in Regenerative Medicine. She co-leads the Cell Technologies research group within the Centre for Biological Engineering of the University, and is currently a member of the Early Career Forum in Manufacturing Research of the EPSRC. The overarching themes of Karen’s research are the manufacture of cellular therapies and the use of cells in the drug discovery process. Karen has a first-class Bachelors degree in pharmacology from the University of Bristol (UK) and a PhD from the Department of Pharmacy and Pharmacology at the University of Bath (UK). Corresponding author:. Coopman, K. (k.coopman@lboro.ac.uk) 922 www.drugdiscoverytoday.com 1359-6446/06/$ - see front matter ß 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.drudis.2013.05.016 Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 Limitations of current 2D systems In vitro cell-based models offer several advantages over most cellfree methods because they do not require purification of the target protein, they immediately select against compounds that are generally cytotoxic or cannot permeate cell membranes, and enable measurement of functional transport activity that is more representative of the physiological state [5]. Using primary human cells in these assays is the ideal gold standard [6]; however, freshly isolated cells, cell cultures or tissue fragments cannot be used to study effects that occur over a longer time period owing to the loss of functional integrity from drug transporter loss and enzyme depletion after a few hours [7,8]. In addition, primary human cells are a scarce and precious resource. For these reasons, immortalised cell lines, such as (transfected) MDCKII-MDR1 or Caco-2, are often used in favour of primary cells. Such cells are relatively inexpensive and easy to propagate, especially in high-throughput systems; however, owing to the genetic transformation process, aspects of the original cell that are important to ADMET studies might be lost, such as transporter expression and function, and cell adhesion molecules [8]. Although relatively simple and effective when using the appropriate cell type, 2D in vitro cell-based assays encounter the traditional problems of static culture: a constantly changing environment owing to nutrient depletion and metabolic waste accumulation. Strategies to improve such systems have been introduced based on mimicry of the natural physiological environment of the organ in question. The re-establishment of heterotypic cell–cell and cell–extracellular matrix interactions through co-cultures and sandwich cultures of cells, respectively, have proven useful in the progression of in vitro model development for several tissue types, including liver [9,10], brain [11] and lung [12]. These models show improved functionality and growth compared with their single culture counterparts, whilst keeping their ease of propagation and use compared with primary cell cultures, which is imperative when developing long-term in vitro models for pharmaceutical applications. Perfusion systems of isolated organs or tissue fragments offer the closest ex vivo model of the natural physiological state, alleviating the problems encountered with static culture, maintaining 3D architecture and prolonging the functionality, heterogeneity, structural complexity and cell–cell interactions [13]. However, the shortage of organs available and the costs and difficulties of maintaining primary cells ex vivo make perfusion cultures unsuitable as a routine in vitro cell-based system. Moving towards 3D systems The development of 3D constructs for pharmaceutical research now encompasses a variety of systems and devices. 3D cell culture with cell types native to a specific organ that are grown on natural scaffolds, such as collagen, decellularised scaffolds or synthetic biodegradable scaffolds, present a bridge between 2D culture and the in vivo cellular environment. 3D devices for pharmaceutical research, with structural complexity and native functionality, are becoming available; however, these cultures are difficult to maintain in vitro and still lack the vascular component of tissues. Extracellular matrix (ECM) coatings that release growth factors and that can be coated onto scaffolds (e.g. polymer scaffolds) have been investigated to aid cell attachment, growth and partially represent the native environments, which can have profound effects on the morphology, behaviour and functions of cells needed in drug discovery processes; however, further work is needed to progress these cultures for use as models [14]. Bioartificial devices combine flow-through systems, usually a form of simple plug flow bioreactor, with cells, thereby mimicking both the dynamic and cell-specific aspects of native tissues [13]. They are being incorporated in the pharmaceutical industry owing to the potential payoff of getting more drugs to market.3 As we discuss below, these devices are being developed from clinically used systems and, although they offer cellular and environmental similarities, most systems currently lack the complex architecture and microenvironments seen in 3D constructs. Status of bioreactors in the development of in vitro models Bioreactors are used in a range of applications, from the production of biopharmaceuticals to applications in tissue engineering, such as cell expansion, generation of 3D tissue constructs and direct organ support devices [15] (Table 1). Unlike 2D static cultures, bioreactors provide a controllable environment in terms of pH, temperature, nutrient supply and shear stress for any cells or cellular constructs incorporated into them. However, key limitations in 3D culture, such as mass transfer of oxygen and nutrients to cells, still apply. For example, cardiac muscle tissue contains a 3 Cuddihy, M. (2011) 3D Cell Cultures Prevalent in Pharma: Human Predictive Functional Tissue Models Conference. Available from: http://3dbiomatrix. com/2011/12/01/3d-cell-cultures-prevalent-in-pharma-human-predictivefunctional-tissue-models-conference/ [Accessed 21 May 2013]. www.drugdiscoverytoday.com 923 Reviews  FOUNDATION REVIEW state, which might be lacking in 2D static cultures. Tissue engineering advances show promise in the development of models with organised structure and cell types similar to in vivo. However the dynamic aspect of in vitro models is yet to fully be addressed, although it is starting to be investigated with the use of 3D bioreactors seeded with cells. Within the pharmaceutical industry, the absorption, distribution, metabolism and excretion properties of NCEs are typically investigated as part of DMPK science. These DMPK properties are optimised within drug discovery so that an NCE has the right clinical profile; for example, ideally, high oral bioavailability, an elimination half-life that enables a once-daily dosing regimen, sufficient exposure at the target tissue and an absence of drug–drug interaction (DDI) potential [4]. In vitro 2D cell-based models are routinely used within pharmaceutical research to investigate the DMPK properties of absorption (Caco-2, derived from human colon carcinoma cells), distribution (MDCKII-MDR1, derived from canine kidney cells) and metabolism and/or excretion (hepatocytes) for NCEs. More recently, some 3D models have started to be investigated to assess their utility in studying aspects of ADMET, namely the bioartificial liver (distribution, metabolism and elimination), the BBB model (distribution) and the bioartificial kidney (excretion). Therefore, in this review, we focus on the potential of bioartificial devices that have been developed for therapeutic use to progress to pharmaceutically relevant in vitro models for use in DMPK studies. REVIEWS Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 REVIEWS TABLE 1 Bioreactor designs and uses in tissue-engineering applications Bioreactor type Reviews  FOUNDATION REVIEW Pros Cons Uses Refs  Induces movement of oxygen and nutrient within culture medium  Easy to scale up  Perception that shear stress can have damaging effects on cells  Poor mass transfer when applying to static scaffolds Growth of cartilage constructs [18]  Efficient surface area for attachment  Potential for immunoisolation  Perception that cells are protected from shear  Nonuniform cell distribution  Fibre membrane can act as a physical transport barrier Generation of a bioartificial liver support system [19]  Uniform cell distribution  Good distribution of nutrients and gas transfer  Stable microenvironment  Complex to scale up  Cells exposed to shear stresses  Low surface area:volume ratio Cartilage construct engineering from chondrocyte-seeded scaffolds [20]  Promotes 3D structure formation  Reduced physical transport barriers  Easy to scale up  Nonuniform distribution of cells  Cells exposed to shear stresses  Nutrient and oxygen transfer issues Early bioartificial liver system studies [21] high density of cells and has a high oxygen demand, which in vivo is supplied by a dense network of capillaries through the tissue. To gain a better understanding of the native environment of cardiac tissue and its influence on the synchronously contracting cardiac constructs, a bioreactor-based model has been developed [16]. Incorporated into this device are rodent cardiomyocytes and fibroblasts grown on scaffolds with a parallel array of perfused channels to mimic the capillary network. Blood is mimicked by supplemented media and the model can be used to assess several parameters, such as the impact of perfusion rates of oxygen on viable cell density, to define the conditions needed to culture cardiac constructs with clinically relevant thickness [17]. Hollow fibre and perfusion systems in particular are widely used in the generation of in vitro models. In addition to enabling users to investigate the impact of environmental conditions on cells, they can also be used to model pharmacological interactions of drugs with cells in infectious [18] or model disease [19] states. Progression from these systems has generated multicompartmental bioreactors that mimic the physiological environment more closely. As well as being used clinically [20], they are being utilised as in vitro models, such as the scaling down of a 3D perfusion multicompartment hollow fibre liver bioreactor for 924 www.drugdiscoverytoday.com use in in vitro pharmacological studies [21]. This bioreactor comprises three interwoven sets of hollow fibre, two for countercurrent medium perfusion and one for gas supply (Fig. 1). Primary human liver cells (both parenchymal and nonparenchymal) are co-cultured in the extra capillary space (ECS), and have been shown to exhibit prolonged cytochrome P450 (CYP) enzyme activity when maintained for over 3 weeks, enabling long-term pharmacological studies to be carried out. Hollow fibre bioreactor (HFB)-based models are also being developed for the BBB and kidney. It is important to assess critically the individual features of established systems so that optimised models for pharmaceutical research can be generated, and this is the focus of the remainder of the review. Perspectives from established systems used within pharmaceutical research Bioartificial liver devices The liver is the main excretory organ within the body, but as well as eliminating waste products through the bile, it also carries out a variety of highly specialised functions, including oxidative detoxification, intermediate metabolism and supply of nutrients, modulation of immune and hormonal systems, and protein and Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 (b) Reviews  FOUNDATION REVIEW (a) REVIEWS (c) 2 ml 8 ml 800 ml Drug Discovery Today FIG. 1 (a) Smallest capillary unit with medium capillaries that are independently perfused (red and blue) and one gas capillary (yellow); cells are cultured within the extracapillary space (cell compartment). (b) The bioreactor comprises three interwoven capillary bundles, each made of multiple hollow fibre capillaries for countercurrent medium perfusion (red and blue) and gas supply (yellow), which enables decentralised nutrient and O2–CO2 exchange with low gradients. (c) Downscaling of the clinical-scale bioreactor prototype with a cell compartment volume of 800 ml was realised by reducing the length and number of capillary layers in two axes of 3D space within the bioreactor, resulting in a laboratory-scale bioreactor with a cell compartment volume of 8 ml and a further downscaled model with a cell compartment volume of 2 ml. Reproduced from [21]. macromolecule synthesis [6]. Disruption in these functions owing to liver disorders, such as acute or chronic liver failure, can lead to a possible 90% mortality rate if essential hepatic functions are not restored during the crucial phase of liver failure [22]. Therefore, there is great emphasis on the development of reliable in vitro models to assess drug interactions and pharmacological responses to prevent drug-induced acute liver toxicity, and to model liver disease states accurately for the development of NCEs targeting disease-induced liver failure. Bioartificial liver (BAL) devices have gone through many iterations while being developed as an extracorporeal device for the treatment of liver failure [23]. Early flat plate and monolayer bioreactors exhibited uniform cell distributions and microenvironments, but presented complex scale-up issues and a low surface area:volume ratio [24,25]. Perfused beds and/or scaffolds promote 3D architecture of seeded cells, which when considering the complex architecture of the liver, is an advantage. The cells showed improved metabolic and synthetic function compared with cells cultured in a spinner flask [26]; however, exposure to shear forces and nonuniform perfusion of medium require an improved bioreactor design. Conventional hollow fibre cartridges offer increased surface area for attachment and immunoisolation of cells when seeded in the ECS [27]. Cells are also protected from shear as medium flows through the fibres. However, nonuniform cell distribution along the fibres is a common problem and, although being beneficial in terms of support and improved cell growth and performance if adapted to the specific cell type [28], membranes can present a physical barrier against the transport of nutrients and metabolites. As mentioned previously, development from standard HFBs to multicompartmental devices, such as that in Fig. 1, has been a natural progression, given the native architecture of the liver, and has been utilised for both clinical and pharmacological applications [21]. Cell sources Hepatocytes are the primary choice for use in BAL systems because they are recognised to be the closest cellular model to the liver [29,30], and carry out most of the in vivo metabolic processes, including the synthesis and excretion of albumin, metabolism of amino acids, urea production and the processing and elimination of drugs and toxins. Phase I and phase II www.drugdiscoverytoday.com 925 REVIEWS Reviews  FOUNDATION REVIEW drug-metabolising enzymes are crucial in the processing of drugs within the liver; therefore, preservation of these enzymes within a model is essential to reproduce cellular environments in vivo. Cell lines have frequently been utilised for in vitro studies owing to prolonged cell survival and ease of maintenance compared with primary hepatocytes. Popular liver-derived cell lines, such as the HepG2/C3A cell line derived from hepatomas, have been used in pharmacological studies, and have recently been utilised in microfluidic bioreactors to examine the sensitivity of the models in response to various xenotoxins [31,32]. However, compared with primary hepatocytes, functionality is often lost or altered in these cell lines [6]. For instance, phase I enzymes were expressed at significantly lower levels in HepG2 cells relative to primary human hepatocytes [33]. Therefore, primary cells would be ideal for incorporation into a pharmacologically relevant in vitro model. However, recent work has demonstrated the successful incorporation of a HepaRG human hepatoma cell line into a 3D BAL device, in which metabolic and transporter functionalities of the cells are preserved from a mere few hours in suspension to at least 1 week in the 3D device, thereby enabling performance of longer-term in vitro pharmacological studies [34–36]. Primary porcine hepatocytes have been widely incorporated into BAL devices [37,38] because they are readily available, are able to maintain differentiated metabolic functions under certain culture conditions, provide a high yield of cells [39] and have a similar metabolic profile to human hepatocytes [40]. However, the environmental triggers that sustain differentiated hepatic function in vitro have not yet been fully characterised [41]. Utilising porcine cells addresses the problem of finding a sustainable source of hepatocytes; however, it would be important to obtain cells derived from a human source to establish a representative model of the natural physiological environment for in vitro pharmacological studies in both normal and diseased states. Primary human adult hepatocytes are considered to be the gold standard for biocompatibility and functionality within BAL systems [6,42], but exhibit a rapid, time-dependent, general loss of function in vitro [7], including reduced cytochrome P450 and other biotransformation enzyme activities [43] and the downregulation of liver-enriched transcription factors, such as CCAAT/ enhancer binding protein alpha [42]. However, the use of complex chemically defined media has enabled hepatocytes to go through several cycles in culture, as well as the stabilisation of hepatocyte morphology, survival and liver-specific functions [37,42,44,45]. Co-culture with nonparenchymal cells has also been shown to preserve liver-specific functions, including initiating physiological reorganisation of cells and the distribution of canalicular transporters similar to the pattern found in vivo [21]. The mechanisms behind these changes are not fully understood, but highly conserved signalling pathways are thought to be involved [9]. There is also limited availability of adult human hepatocytes owing to their acquisition from liver biopsies, which has presented problems with utilising this source in clinically used devices because the number of cells needed for seeding a BAL for the treatment of hepatic failure is large, approximately 2  1010 [25,46]. However, the number of cells needed for bioreactor-based models is approximately 3  108 per device [21], which could be attained with minimal culture in vitro. 926 www.drugdiscoverytoday.com Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 Device designs and functionality Within the native liver, hepatocytes are ordered in a 3D lobular network of hexagonal constructs, surrounding a central blood vessel. Individual cells have polarity between the apical domain interfacing with bile canaliculi and a basal domain interfacing with the sinusoid, helping localise specific functions. During culture, hepatocytes flatten and spread upon attachment, whether grown on a flat surface or a scaffold. Although cell–cell junctions can be formed, the cytoskeleton is largely disrupted, and leads to loss of cell polarity and, ultimately, liver-specific functions [6]. Nonetheless, C3A cells and primary porcine cells have been utilised within HFBs in clinically tested devices. The extracorporeal liver assist device (ELAD) comprises a dual pump dialysis system connected to one or more hollow fibre cartridges seeded with C3A cells in the ECS. Blood is ultrafiltered and pumped through fibres, with plasma flowing through the ECS. This direct interaction enables cellular processes to occur before the ultrafiltrate is filtered by a dual membrane to remove cells and cellular debris and returned back into the blood stream [47]. The liver has an oxygen-rich environment, fed by the hepatic arteries and portal vein. Unsurprisingly, an improvement in functionality of C3A cells in HFBs was noted when red blood cells were added in culture medium [48], highlighting that an adequate oxygen supply is a prerequisite for the maintenance of hepatocytes in vitro. Considering this, the Excorp medical bioartificial liver support system (BLSS) incorporates an oxygenator and heat exchanger along with a blood pump and HFB seeded with primary porcine hepatocytes in the ECS [49]. Whole blood flows through the fibre, protecting the cells from host immunological responses [38]. Another BAL, the modular extracorporeal liver support system (MELS), has attempted to take into account the biliary system of the liver, which excretes conjugated bilirubin and metabolites. Hollow fibre bundles, seeded with human hepatocytes from discarded donor livers, are incorporated alongside a detoxification module for removing albumin bound toxins. This provides an inlet of plasma and contact with hepatocytes in the ECS, an outlet of plasma, and hydrophobic membrane bundles for gas exchange inside the bioreactor [50]. The structure of the seeded bioreactor module is an early iteration of the multicompartmental bioreactor being developed for in vitro pharmacological studies by Gerlach et al. [6] (described above and in Fig. 1). These systems, although showing positive biochemical changes in patients and maintaining patients through to transplantation, show no significant effect on survival and, therefore, can be assumed to be unsuitable replicates of the environment in vivo. Aspects such as co-culture of nonparenchymal and parenchymal cells, efficient oxygenation of cells and the presence of a biliary system for the removal of toxic protein-bound substances and other metabolites should be taken forward in the generation of both efficient in vitro models, as has been demonstrated in the multicompartmental liver bioreactor (Fig. 1), and clinically relevant medical devices. Use of BALs in DMPK studies Bioreactor-based models are currently being integrated into the DMPK process, initially assessing the functionality of cells within the systems relative to drug transporter and metabolic enzyme expression and function. When investigating the metabolism of Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 For both the medical devices and in vitro models highlighted above, there are currently no standardised criteria to define the efficiency and efficacy of these systems. Furthermore, any criteria applicable to a device will be partly dependent on the type of bioreactor used (i.e. hollow fibre systems require tight monolayers of cells on the extra luminal side of the fibre membrane). However, cellular metabolic functions and enzymatic and transporter retention are paramount in all bioartificial liver systems. From the clinically relevant devices mentioned above, there is little focus on the performance of cells, and factors used to determine the efficiency of the systems before being used in clinical studies are not clearly stated. Devices that have been utilised for DMPK research go further in defining the criteria used to assess the models, such as monitoring metabolic activity via albumin synthesis, lactate dehydrogenase or aspartate aminotransferase, and drug-metabolising enzymatic activity via administration of drugs with cytochrome P450-dependent metabolism, as well as drug transporter functionality and visualisation by inhibitor transport assays and staining [21,34–36,51–53]. With the progression of these systems, both medical and in vitro models, standardised criteria defining the performance relative to native conditions of the above-mentioned factors should be produced. BBB models The BBB is one of the most important lines of defence against infections in the brain, acting as a physical and metabolic barrier between the central nervous system (CNS) and systemic circulation and maintaining homeostasis within the brain. The capillary structure is distinct from that found in peripheral tissues, with the presence of tight junction proteins between endothelial cells (ECs), the restriction of transcellular movement of molecules and contact of ECs with astrocytes to provide support [54], protection from hypoxia and aglycaemia and to separate capillaries from neurons [55]. The decreased paracellular permeability of the BBB, while having multidrug transporters such as P-glycoprotein (P-gp, also known as MDR1), prohibits the transport of large drug molecules into the CNS, presenting the need for a reliable in vitro model for ADME studies. Existing models have not been characterised as well as liver devices, partly owing to their lack of clinical relevance regarding the generation of an extracorporeal device, but also owing to the complexity of cell alignments and functionality within the model. However, similar to the BAL devices, HFBs have been developed and utilised as ‘dynamic’ models of the BBB [56]. Model designs Unlike liver models and bioartificial devices, there is no one cell type that can be incorporated to mimic the BBB. Most models reported in the literature use two cell types; ECs and glial cells (astrocytes). Early models simply used mono- and/or co-cultures of ECs alone or with glial cells cultured on transwell plates. This approach is attractive owing to its simplicity; however, primary ECs lose their BBB properties in vitro owing to the absence of stimuli that are normally present in vivo [56]. Furthermore, specific transporters, such as P-glycoprotein (P-gp), are downregulated in the absence of astrocyte-derived soluble factors, leading to abnormal permeability across the monolayer [57]. This dedifferentiating process might also be accelerated by nonphysiological conditions, www.drugdiscoverytoday.com 927 Reviews  FOUNDATION REVIEW diclofenac, freshly isolated human hepatocytes were found to produce 90% less hydroxydiclofenac, the hydroxylated product of diclofenac generated directly by the cytochrome P450 enzyme CYP2C9, after a 3-day culture in 2D conditions compared with a high level of metabolites from human hepatocytes in the multicompartmental bioreactor described above [21,34]. The major intermediary diclofenac metabolite, diclofenac acyl glucuronide [produced from the glucuronidation of diclofenac by UDP-glucuronosyltransferase (UGT) enzymes before hydroxylation by the P450 enzyme CYP2C8], was found to be absent in human hepatocyte suspensions, but present in the hepatocyte-seeded bioreactor, possibly owing to the bioreactor system mimicking the in vivo environment and enabling the efflux of metabolites away from the cells, thereby avoiding further metabolism. Overall, the relative levels of the metabolites found in the suspensions versus the 3D cultures were similar; however, all human in vivo metabolites were detected in the bioreactor culture, whereas one major in vivo metabolite was not present in the hepatocyte suspension. In addition, the biotransformation capacity of the hepatocytes within the bioreactor was retained for at least 1 week, providing a good model for metabolite investigations from slowly metabolised drugs [34]. The activity of CYP3A4 was demonstrated in the 3D bioreactor system by monitoring the metabolism of atorvastatin acid (ATA) to its 2- and 4-hyroxylated metabolites [35], and for HepaRG cells in the 3D system by monitoring midazolam, which indicated that CYP3A4 activity was retained for up to 2 weeks [47]. This retention of CYP3A4 activity is longer than that currently available in other in vitro models. Transporter expression is also a key aspect in the functionality of hepatocytes for DMPK screens. Within the bioreactor, immunohistochemical staining of multidrug resistance-associated protein (MRP2) and multidrug resistance protein (MDR1), both bile canicular efflux transporters, showed expression, localisation and distribution similar to native liver tissue for at least 2 weeks within the bioreactor [21,36]. Expression of the basolateral uptake transporter organic anion transporting polypeptide 1B1 (OATP1B1) has also been investigated on hepatocytes within the 3D system preserved for 9 days, functional activity of OATP1B1 was demonstrated by the transport of substrates estradiol- 17b-D- glucuronide (E17bG) and atorvastatin acid (ATA) in the presence and absence of the inhibitor estrone-3-sulphate (E3S), and was retained for 7 days compared with the decrease of functional activity of OATP1B1/ 1B3 in plated primary human hepatocytes after just 2 h [35,51]. When assessing drug–drug interactions, OATP1B1 is a vital hepatic transporter to evaluate, because it is inhibited by many commonly used drugs, and is responsible for the targeted uptake of the widely prescribed statins to their site of action [52]. This has been demonstrated by similar increases in the elimination half-life of ATA in the presence and absence of OATP1B1 inhibitors, with a 2.7-fold increase in the presence of rifampin [53], and a 1.7-fold increase in the presence of E3S within the bioreactor model. The difference in these values had been attributed to experimental setup and potencies of the two inhibitors, and it is also been noted that the major limitation of this model lies in the variable quality of isolated human hepatocytes from different donors [35]. However, these initial studies into the functionality of drug transporters and metabolic enzymes clearly establish the potential of this model for use throughout the DMPK process. REVIEWS Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 REVIEWS TABLE 2 Performance of various cell types within the HFB in the generation of an in vitro BBB model Reviews  FOUNDATION REVIEW EC Glial cell Days prolonged in culture Refs Rat brain microvascular ECs Rat brain astrocytes 30 [62] Human brain ECs Human foetal astrocytes Not reported [62] Bovine aortic ECs C6 glial cell line 30 [62] such as the presence of serum on both sides of the polarised EC monolayer instead of only on the luminal surface in vivo. Coculture models incorporating glial cells mimic the in vivo situation more closely, demonstrated by the increased expression of brain endothelial marker enzymes, transporters, and an increased transendothelial electrical resistance (TEER) across the barrier indicative of a ‘tight’ monolayer. This model is effective for studies around the function of the BBB and processes and interactions between the ECs and the glia [56]. However, for a replicative model of the BBB, factors such as shear stress, which is vital to promote growth inhibition and differentiation of ECs and enables the trafficking of metabolic fuels to the brain, should be incorporated into animal-based models before ideally progressing into humanbased models, to enable direct extrapolation of data from studies, and functional and physiological mimicry of the human BBB in vivo [58]. Dynamic (flow-based) models incorporate HFBs with coated fibres seeded with ECs (intraluminally) and glial cells (extraluminally). In an early model developed by Stanness et al. [59], ECs and corresponding glial cells from human, rat and bovine sources were cultured successfully in a HFB coated with ProNectin (Table 2) [59]. Morphology and growth patterns were consistent with all cell types; however, it was noted that when differing types of EC and glial cells were cultured, such as human umbilical cord ECs with C6 cell line glial cells, BBB properties were not developed. Poor EC growth in the absence of glial cells within the hollow fibres was observed via low transendothelial resistances and studies assessing the transport of morphine demonstrated permeability ratios to be similar to in vivo, but absolute values to be lower. This was thought to be because of the differences in flow rate patterns in the devices and from in situ. The results supported the use of HFBs in BBB models; however, because bovine aortic ECs + C6 seeded cartridges were used for drug transport studies, utilising human brain ECs in the HFBs would be a natural progression. Neuhaus et al. [60] utilised a commercially available bioreactor from the same manufacturer as Stanness et al., and seeded immortalised porcine brain microvascular endothelial cells (PBMEC/C12) intraluminally with the glial cell line C6 cells in the extra capillary space. Medium was pumped through the system applying shear stresses of 2.7–3.9 dyn/m2, mimicking conditions in brain capillaries. Results of cell growth and transport of benzodiazepines were compared between the transwell and flow-based models. The results showed that PBMEC/C1-2 cells seeded on the transwell inserts started detaching after 5 days, prohibiting the use of transwell systems in long-term experiments, whereas growth and morphology of the ECs in the HFB was affected owing to exposure to flow, prolonging the survival of the cells within the device. Permeability coefficients of benzodiazepines were also 928 www.drugdiscoverytoday.com shown to be lower in the HFB model, presumably owing to tighter monolayers and improved barrier functions within the device [60]. The BBB is an active site in terms of transport, expressing multiple drug transporters to restrict compounds with high affinity for efflux transporters and low transcellular membrane permeability into the CNS. However, owing to its complexity, there are few established in vitro models, and CNS-targeted compound analysis relies heavily on in vivo and in silico assessments. P-gp (MDR1) is a predominant efflux transporter in the BBB, capable of limiting the penetration of a wide range of substrates into the brain. For this reason, MDCK–MDR1 cellular bidirectional transport assays are currently used in DMPK screening to assess both CNS-penetrating and CNS-restricted compounds, as appropriate depending upon target profile [4]. Differences in transporter expression and passive permeability have been shown in 2D static co-culture models (described above). However, investigations have also highlighted interspecies variability, with a twofold higher permeability in a rat EC and glial model than in the human counterpart. The limitations of the current transwell model utilising transfected cell lines are its static nature and lack of availability of human cells. However, this system could be improved by using rat ECs and glial cells, to better represent the in vivo environment than is currently available with transfected cell lines alone, to account for the difference in permeability [61]. Owing to the nature of individual transporter testing in current 2D cultures, 3D models exhibiting BBB characteristics could also be utilised for drug–drug interaction studies as the 3D liver models are starting to investigate. Overall, the outcomes of these models are promising for the use of HFBs in pharmacological applications. However, the human brain is more complex than its rodent counterpart and, although many drug transporter proteins are expressed in rodent ECs [56], primary human ECs and glial cells would ideally be incorporated into an optimised device. This might be difficult to generate for large-scale studies because healthy human glial and ECs are rarely available, so one possible source would be the differentiation of stem cells into these cell types. However, extensive work in stem cell differentiation and characterisation from embryonic and induced pluripotent stem cells is needed, and is not a viable source for models at present. Recreating aspects of the microenvironment of the CNS in terms of ECM proteins surrounding glial cells might also improve performance of the cells within the device. Progressions in bioartificial kidney (BAK) devices as an in vitro model The need for an optimised model from existing clinical devices The kidneys are primarily responsible for the excretion of waste products, such as urea and ammonia, from the body through the Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 (a) REVIEWS (b) 200ml/min V A A V Continuous haemofilter 150ml/min 10ml/min Reviews  FOUNDATION REVIEW 4ml/min Waste 185ml/min 7ml/min 10ml/min Haemofiltration cartridge Bioartificial cartridge Bioartificial cartridge (RAD) 3ml/min 5ml/min Waste Waste Drug Discovery Today FIG. 2 Schematic diagram of (a) bioartificial kidney device used for the treatment of acute renal failure developed by Humes et al. [66]. (b) Bioartificial kidney device developed as a renal replacement therapy for chronic renal failure by Saito et al. [65]. The devices have similar components and configurations; however the device developed by Saito et al. incorporates a continuous haemofilter for continuous treatment. Based on device layouts from [65,66]. urine. They also provide a variety of significant secondary functions through homeostatic, endocrine, immune and metabolic processes. Chronic kidney disease (CKD) is a growing problem within the UK, with approximately 1.8 million people with CKD in England alone. This is putting a growing burden on the NHS, costing over £1.4 billion per year, which is largely because of renal replacement therapy (e.g. dialysis), which showed a 29% increase in usage from 2002 to 2008, and complications, such as heart disease and stroke [62]. This highlights the need for early diagnosis of the disease, more efficient and cost-effective renal replacement therapies, more effective drug treatments and, consequently, an optimised kidney model for pharmacological studies. Existing bioartificial kidney (BAK) devices have been developed primarily as a renal replacement therapy. They incorporate both mechanical and biological processes, pumping haemofiltered blood through a renal assist device (RAD). Coined by Humes et al., this device contains renal cells seeded in the lumen of hollow fibres of a standard haemofiltration cartridge [63], contrary to BALs, which utilise cells within the ECS. Unlike dialysis, which provides intermittent filtration, in vitro studies of the BAK device show differentiated active transport of essential nutrients, such as glucose, amino acids and sodium bicarbonate, sufficient metabolic activity and important endocrine processes [64]. Although the current bioartificial systems have been utilised for clinical use, there is scope as with the systems mentioned previously to use them as a foundation for many other applications. There have been two main groups focusing on the development of a BAK device for clinical use. Humes et al. developed a device to treat patients with acute renal failure and multiple organ failure with a degree of success, whereas Saito et al. focused on the prevention and treatment of long-term complications in patients on maintenance dialysis [65]. However, both devices utilise the same bioartificial component, hollow fibre modules seeded with proximal tubule cells (PTCs), and pass ultrafiltered blood through the module before administering back into the patient (Fig. 2). Humes et al. [66] utilised commercially available haemofiltration cartridges, with polysulfone (PSF)-based fibres seeded with human PTCs (hPTCs). Although confluent monolayers of cells were observed within the fibres [64], Saito et al., concurring with Oo et al. [67], have commented on the lack of biocompatibility of PSF and PSF-based membranes for generating a confluent monolayer of cells [65]. The clinical trial undertaken with the BAK device from Humes’ group has also been critically assessed, noting the lack of documentation on expected effect size, the incompletion of treatment of patients assigned to continuous veno-venous haemofiltration(CVVH) + renal assist device (RAD) treatment and the lack of statistical significance of the primary results with comparisons performed as an as-treated rather than an as-randomised intention-to-treat sample [68]. Cell sources To create an efficient, reliable model of the kidney in situ, a BAK device must incorporate a cell type that can carry out several native functions; therefore, research has been drawn to renal PTCs. These cells perform a variety of renal-specific functions, including reabsorption, metabolic and transport functions, the secretion of uraemic toxins and xenobiotics, and performing immunomodulatory functions [69]. Renal PTCs also have a range of transporters, such as organic anion and organic cation transporters for basolateral drug uptake and ATP binding cassette (ABC) transporters present on the apical surface for luminal excretion into urine. Many of the transporters are polyspecific; that is, they accept www.drugdiscoverytoday.com 929 REVIEWS Reviews  FOUNDATION REVIEW compounds of different sizes and molecular structures, and have overlapping substrate specificities [70]. Primary hPTCs are required for clinical BAK devices, but the sheer number of cells needed (approximately 109 renal cells per device) is a major obstacle to their widespread use, driving the search for alternative cell sources. Human PTCs have also been used as a wellcharacterised in vitro model of the kidney in drug transport studies, owing to the classification of specific carrier proteins on the plasma membrane by substrate specificity and functional assessment, and the prediction of drug–drug interactions based on competitive inhibition for transport [71]. Two transporters, organic anion transporter (OAT) 1 and OAT3, are thought to be the major OATs responsible for the basolateral uptake of various organic ions from the blood in vivo; OAT3 action was demonstrated by the mediation of rosuvastatin uptake in vitro in the presence and absence of probenecid, a uricosuric drug [72]. Probenecid has also been shown to be a potent inhibitor of OAT2 and OAT4, where prostaglandin F2a (PGF2a) uptake by OAT2, and estrone-3-sulfate uptake by apically located OAT4, were inhibited in the presence of probenecid. The main uptake transporter of organic cations in hPTCs is thought to be OCT2, located on the basolateral membrane. Functionality has been demonstrated in hPTCs by the decrease in uptake of a fluorescent dye, ASP+ in the presence and absence of the inhibitor quinine [73]. Although drug–drug interactions for organic cations have been observed in other cell sources, such as LLC-PK1 cells, few data on the drug interactions for OCT2 have been documented [74]. Similar to liver, P-gp is documented to transport a wide range of xenobiotics in hPTCs, in particular anticancer agents, such as methotrexate and cisplatin [75]. However, in hydrophilic statin efflux, P-gp was found to have a minor role compared with the MRP2 and breast cancer resistance protein (BCRP) efflux transporters [72]. Differences between effects of drugs in vivo and in vitro have also been highlighted by the inhibitory effects of KW-3902 and betamipron on OAT4 in vitro, but that showed no significant inhibitory effects on OAT4 in vivo [76]. The expression of both phase I [11 CYP enzymes and three glutathione S-transferases (GSTs)] and phase II (three UGT and variable expressions of three sulfotransferases) drug-metabolising enzymes has also been found to be retained within primary hPTC cultures. Furthermore, when grown in a confluent monolayer, which is necessary in current BAK devices, the expression of these enzymes was generally maintained at a measurable level, albeit lower than in the native tissue [77]. Selective drug transporter loss upon primary cell culture has been identified to be a consequence of dedifferentiation. However, recent novel human tubular kidney cell cultures grown on permeable filter supports have been shown to express a wide range of drug transporters owing to the co-culture of proximal and distal tubular cells [70]. This finding is concurrent with the increased function from co-culture of hepatocytes and nonparenchymal cells in BAL devices. Collectively, these results indicate that primary cultures of hPTCs can metabolise drugs, provide good models for drug interaction studies and benefit the clinical setting when incorporated into BAK devices. Although hPTCs do not always represent in vivo functionality, these results could be improved by their incorporation into a dynamic bioreactor, providing a more accurate model of hPTCs in situ. Using immortalised cell lines eliminates the problem of needing vast numbers of cells for studies. Methods such as cloning, 930 www.drugdiscoverytoday.com Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 derivation from cancerous cells and viral transfection can all produce genetically stable, but functionally altered cells. Human cells are relatively difficult to immortalise and do not express differentiated features in vitro [64,78]. Therefore, the established cell line HK2, generated from the infection of human PTCs with an immortalising construct from the human papilloma virus, and the more recently developed human kidney proximal tubular cell line (HKC) from exposure of renal epithelial cells to Adeno 12:SV40 hybrid virus, do not provide an adequate replacement for primary cells for use in clinical BAK devices [78]. Other established animalderived cell lines, such as LLC-RK1, LLC-PK1 and MDCKII, have also provided adequate models for research, but interpretation to the human situation is often difficult owing to species differences, especially with regard to transporters and enzymes. Drug transport with human renal cell lines has not been widely researched, possibly owing to the substantial loss of differentiated function from the immortalisation of human cells, and primary hPTCs or transfected animal cell lines, such as MDCKII-MDR1, are more frequently used. This highlights the challenge of generating a physiologically complex, well-characterised and validated in vitro cellular model that can replace current US Food and Drug Administration (FDA)-recommended cell transport test systems, such as P-gp transfected MDCK or LLC-PK1 cells [79]. Porcine PTCs are widely used in kidney research owing to their reliability, and anatomical and physiological similarity to hPTCs. In preclinical studies, porcine PTCs in a BAK device were shown to tolerate the uraemic environment while providing metabolic, reabsorptive and endocrinological activity in uraemic dogs [80], enabling research to further investigate the benefits of BAKs, and progress to hPTC utilisation for clinical trials. There are few studies of the differences in drug interactions and metabolism between porcine and human PTCs to date, although drug transporter expression on primary porcine PTCs and cell lines are comparative to native human kidney cells (Table 3). Both primary and porcine cell lines are readily available and are an established model for drug toxicity; however, an optimised model for drug toxicity would require human PTCs instead of the readily available established porcine model. Owing to their multipotency and self-renewal capabilities, embryonic stem cells (ESCs) might provide a reliable source of PTCs for use in BAK devices when exposed to the right environmental cues [63]. Studies into human ESCs differentiating into renal cells have not been widely undertaken; however, preliminary studies have shown integration of undifferentiated and mesoderm-derived ESCs into the stroma and developing nephrons [81]. They have also been shown to express a wide range of early renal markers, and exhibit upregulation of kidney precursor markers, such as EYA1, LIM1 and CD24 [82]. However, the generation of the different types of renal cell from human ESCs still needs to be investigated thoroughly [81]. Owing to the sparse studies and various obstacles, drug metabolism and transport by human ESCderived renal cells have not been documented, but will be essential in any progression of stem cell utilisation towards potential in vitro cell models used in pharmaceutical research. Device designs As mentioned above, the existing devices used for clinical applications utilise haemofiltration cartridges seeded with hPTCs. Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 REVIEWS TABLE 3 Transporter family Gene product Location MDCK II LLC-PK1 Primary porcine cells Refs Organic anion transporters OAT1 (SLC22A6) Basolateral X X  [84–86]c ABC transporters Organic cation transporters OAT2 (SLC22A7) Basolateral OAT3 (SLC22A8) Basolateral X X H [84,85] [84–86]c OATP1A2 (SLCO1A2) Apical X X  c,d MRP1 (ABCC1); MRP5 (ABCC5) Basolateral H H H [84–86]c,d MRP2 (ABCC2); MRP4 (ABCC4) Apical H H H [84–86]c,d MRP3 (ABCC3) Basolateral H [84,85] MRP6 (ABCC6) Basolateral H [84,85] MDR1 (ABCB1) Apical H H H [84–86]c,d BCRP (ABCG2) Apical H H X [84] X X  [84–86] OCT1 (SLC22A1) Basolateral OCT2 (SLC22A2) Basolateral [84–86] OCT3 (SLC22A3) Basolateral [84–86] Carnitine and/or organic cation transporters OCTN1 (SLC22A4) Apical [84–86] OCTN2 (SLC22A5) Apical [84–86] Organic anion transporting polypeptides OATP4C1 (SLCO4C1) Basolateral [86]d Peptide transporters PEPT1 (SLC5A1) Apical X H H [84–86]c PEPT2 (SLC5A2) Apical X X H [84–86] MATE1 (SLC47A1) Apical [86] MATE2-K (SLC47A2) Apical [86] Multidrug and toxic compound extrusion transporters a Human drug transporter orthologues present on the cell lines and primary cells. Blank fields indicate no presence of transporters on the cells. b Abbreviations: H, present; X, absent; , only detectable after 35 cycles of PCR; , only present in freshly isolated samples. c Jin Jang, K., Ingber, D. (2011) Human kidney proximal tubule-on-a-chip for drug transporter studies and nephrotoxicity assessment. 15th International Conference on Miniaturized Systems for Chemistry and Life Sciences, 2–6 October, 2011, Seattle, Washington, USA (http://www.rsc.org/images/LOC/2011/PDFs/Papers/502_0844.pdf ) [Accessed 21 May 2013]. d FDA. Drug Interactions & Labeling: Drug Development and Drug Interactions: Table of Substrates, Inhibitors and Inducers. Available from: http://www.fda.gov/Drugs/ DevelopmentApprovalProcess/DevelopmentResources/DrugInteractionsLabeling/ucm093664.htm#major [Accessed 21 May 2013]. However, for an in vitro model, device design could be optimised. For example, the number of cells needed could be greatly decreased owing to the reduction in the size of the cartridge. Hollow fibre cartridges could also be greatly optimised, because commercially available modules have varying degrees of biocompatibility. Tailoring fibre materials to different cell types and sources has been thought to improve cell performance and, for primary hPTCs, it is thought that some of the membrane properties needed include a hydrophilic, negatively charged adhesive surface [28]. When a variety of commercially used membranes, such as regenerated cellulose (RC), PSF and/or PVP and polyethersulfone (PES) and/or polyvinylpyrrolidone (PVP) coated with ECM, were tested with primary hPTCs, performance of the cells was not improved, indicating that these materials are not suitable for use in BAK devices. The differing properties between cell sources are further highlighted by reports that PSF membranes coated with ECM are able to maintain both primary porcine [83] and LLC–PK1 cell monolayers for 3 weeks after confluency [84]. Coating commercially available fibres with tailored ECM is preferable to tailoring membrane materials owing to the ease of assembly and cost effectiveness. Various single and double coatings have been tested, including collagen IV and laminin blends with a variety of growth factors [69]. Single coatings demonstrated little improvement on hPTC performance within the fibre, whereas double coatings of 3,4-dihydroxy-L-phenylalanine (DOPA) and collagen IV on haemocompatible membranes revealed functional epithelia formed by hPTCs in a HFB [67]. Bioreactor configuration should also be taken into consideration to both achieve native mimicry and produce an optimal environment for the cells to proliferate. Previous bioartificial devices utilising hollow fibre cartridges have encountered problems with the nonuniform distribution of cells and the lack of transport of gases and nutrients [85]; therefore, including an oxygenator or aerated medium would need to be considered. Additionally, to mimic the kidney, the device would need to include a tubule (hollow fibre) and capillary and/or blood (extracapillary space) section with shear rates according to native conditions. These would be higher in the ECS (typically 10 dyn/cm2 on endothelial cells in vivo) than in the fibre (typically 1 dyn/cm2 on PTCs in vivo) [86]. This could lead to movement of molecules through the membrane if the monolayer of cells was not tight, and affect cell performance owing to varying pressures. However, from previous devices used clincally, no loss of membrane integrity has been reported when using different flow rates in the tubule and ECS [66]. Emerging in vitro models have also adhered to the established hollow fibre structure. Researchers at the Wyss Institute, Harvard University have introduced ‘organs on a chip’. These microfluidic www.drugdiscoverytoday.com 931 Reviews  FOUNDATION REVIEW Drug transporter orthologues present on MDCKII and LLC-PK1 cell lines and primary porcine PTCsa,b Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 REVIEWS PDMS channel Porous membrane PDMS reservoir Reviews  FOUNDATION REVIEW 35-mm culture dish Kidney tubular cells apposed to porous membrane Drug Discovery Today FIG. 3 Schematic of human kidney proximal tubule-on-a-chip in the form of a multilayer microfluidic device that integrates a polydimethylsiloxane (PDMS) microfluidic channel, a porous membrane and a PDMS reservoir. Human primary proximal tubule cells were cultured on the device. Modified, with permission, from Jin Jang et al.4 devices, in this case ‘kidney on a chip’, utilise hPTCs seeded within a microfluidic channel apposed to a polyester membrane. The channel is immersed in medium, and a shear stress of 0.2 dyn/cm2 is applied through the channel (Fig. 3). Cytoskeletal proteins, tight junction proteins, ion transporters and specific drug transporters (e.g. OCT2) have shown improved results in a fluidic model compared with a static version, and nephrotoxic compounds, such as cisplatin, have shown the ability of the microsystem to recapitulate the in vivo toxicity of the drugs. Overall, the system presents a substantial answer to the need of a reliable 3D in vitro model.4 However, functional characterisation of the drug transporters within the device that are acknowledged to be of clinical importance to drug disposition by the FDA, namely P-gp, BCRP, OCT2, OAT1 and OAT3, should be undertaken to establish an ideal model system.5 Concluding remarks and future perspectives Bioreactors have been utilised in many different areas of industry and research and, for the progression of in vitro models of 2D cultures and suspensions to 3D constructs mimicking the natural physiological state in situ, they have proven to be crucial. The generation of bioartificial devices which provide both the structural and biological aspects, needed for cell growth and function, have been exploited for clinical and research purposes. The most established bioartificial system, the bioartificial liver, is a prime example of the development of a bioreactor designed initially for therapeutic use, being utilised as an in vitro model. From simple perfused bed and HFBs, focus on the native liver structure and 4 Jin Jang, K. and Ingber, D. (2011) Human kidney proximal tubule-on-a-chip for drug transporter studies and nephrotoxicity assessment. 15th International Conference on Miniaturized Systems for Chemistry and Life Sciences, 2–6 October, 2011, Seattle, Washington, USA (http://www.rsc.org/images/LOC/ 2011/PDFs/Papers/502_0844.pdf ) [Accessed 21 May 2013]. 5 FDA. Drug Interactions & Labeling: Drug Development and Drug Interactions: Table of Substrates, Inhibitors and Inducers. Available from: http:// www.fda.gov/Drugs/DevelopmentApprovalProcess/DevelopmentResources/ DrugInteractionsLabeling/ucm093664.htm#major [Accessed 21 May 2013]. 932 www.drugdiscoverytoday.com cellular functionality within the device has prompted multifibre bundle designs. In addition to bioreactor design, development of the BBB models has highlighted the importance of cell sources, especially with co-culture of different cell types, within a fluidic system opposed to a static culture. It is understood that primary human cells are the gold standard for an in vitro model; however, owing to the difficulty of acquisition, BBB models have had to utilise other primary sources. Embryonic stem cells could be a possible solution to the dilemma of availability for the future; however, more research is needed to investigate the difference in properties between primitive and mature human CNS and ECs before incorporation into a device for DMPK investigation. Bioartificial kidney devices have followed the progression of BAL systems. From development for clinical use as a renal replacement therapy for acute and chronic renal conditions, they are now starting to reveal their potential as in vitro models (Fig. 4). Primary human cells have been utilised within the system, and are relatively easy to acquire and maintain, especially with the reduced cell number requirement in a model. However, co-culture with distal tubule cells has not been fully investigated within a BAK device, and functional characterisation of drug transporters crucial for ADMET studies have not yet been undertaken in the existing microfluidic system. To date, there are no universal criteria laid out to aid the development of an acceptable and/or suitable bioreactor. However, it is widely accepted that in vivo micro environmental aspects, such as cell viability, cell–cell interactions, cell–matrix interactions, tissue architecture and cell oxygenation, are vital to give an accurate representation of the physiological state and, consequently, the effects of new chemical entities [3]. Therefore, it is important to bear in mind those factors that are specific to the desired tissue when developing in vitro models for use in pharmaceutical research. Furthermore, it is noticeable that little or no measurements are currently taken from these bioreactor-based devices to help control or monitor cell health or functionality during use. Thus, we anticipate that, with the continuing advances being made in sensor technology, the next generation Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 Reviews  FOUNDATION REVIEW REVIEWS Functional cell type Microenvironment Progression of bioreactors as in vitro models Cell type and sources Primary cells, cell lines Cell characterisation Membrane material and properties Transporter,expression, metabolic enzyme possession, native morphology, adherent protein expression, cell functionality Material, pore size, porosity, surface charge, hydrophilicity, topography, shear rate Functional cell support Coatings Similarity to native microenvironment, number of coatings, mixture of coatings Co-culture of cells Bioartificial Bioartificial kidney BBB Bioartificial liver Future developments- Utilising stem cell derived specialised cells Primary human hepatocytes Transporters:MRP2, MDR1, BCRP Metabolic enzymes: CYP1A, CYP2C9 CYP3A Cell functionality markers: Albumin synthesis, lactate dehydrogenase, aspartate aminotransferase, glucose secretion, lactate production Co-culture with non-parenchymal cells PBMEC/C1 (porcine brain microvascular endothelial cells), C6 (rat glioma cells) Glucose consumption, lactate production, paracellular transport, benzodiazapene permeation Human proximal tubule cells γ-glutamyl transferase activity, Future developments- Tools for real time monitoring of cells incorporated e.g. gas/ electrolyte and metabolite sensors Multicompartment hollow fibre bioreactor Native architecture Device type Hollow fibre and perfusion systems Device setup Single circuit, multi-compartmental devices Future developments- Miniaturisation of devices Polyethersulfone, hydrophobic multilaminate hollow fibres Commercially available hollow fibre cartridge Polypropylene, 0.5 µm pores, 4-50 ml/min Fibronectin Single fibre hollow fibre bioreactor Polyethersulfone/ polyvinylpyrrolidone, 0.5 µm pores, 80 µl/min Double coating of DOPA and human collagen IV parathyroid hormone, RT-PCR, SEM, immunostaining Drug Discovery Today FIG. 4 Key aspects from the progression of bioreactor-based in vitro models and future developments [21,60,67]. Abbreviations: BBB, blood–brain barrier; BRCP, breast cancer resistance protein; CYP, cytochrome P450; DOPA, 3,4-dihydroxy-L-phenylalanine; MDR1, multidrug resistance protein; MRP2, multidrug resistanceassociated protein. of devices will incorporate these to improve either device function or longevity. Bioartificial systems have started to show great promise in the generation of in vitro models for various organs and disease states. Observations from existing systems are the importance of mimicry of both structure and environments in situ in the bioreactor, the choice of cell and source to model the functions of the organ and essential factors for maintaining cellular function, such as oxygen availability and nutrient and metabolite delivery. With these factors in mind, successful models utilising bioreactors can be achieved in the near future and will hopefully provide a better means to predict clinical outcome from in vitro data. References 1 Light, D.W. (2011) Drug R&D costs questioned. Genet. Eng. Biotechnol. News 31, 6–7 2 Scherer, F. (2011) R&D Costs and Productivity in Biopharmaceuticals. F Kennedy School of Government, Harvard University 3 Astashkina, A. et al. (2012) A critical evaluation of in vitro cell culture models for high-throughput drug screening and toxicity. Pharmacol. Ther. 134, 82–106 4 Ballard, P. et al. (2012) The right compound in the right assay at the right time: an integrated discovery DMPK strategy. Drug Metab. Rev. 44, 224–252 5 Barberis, A. (2002). Cell-based high-throughput screens for drug discovery. Eur. Biopharm. Rev. Winter, 93–96 6 Gerlach, J.C. et al. (2008) Bioartificial liver systems: why, what, whither? Regen. Med. 3, 575–595 7 Fahrig, R. et al. (1998) Use of primary rat and human hepatocyte sandwich cultures for activation of indirect carcinogens: monitoring of DNA strand breaks and gene mutations in co-cultured cells. Toxicol. In Vitro 12, 431–444 8 Cai, H. et al. (2009) Assessment of the renal toxicity of novel anti-inflammatory compounds using cynomolgus monkey and human kidney cells. Toxicology 258, 56–63 9 Bhatia, S.N. et al. (1999) Effect of cell–cell interactions in preservation of cellular phenotype: cocultivation of hepatocytes and nonparenchymal cells. FASEB J. 13, 1883–1900 10 Swift, B. et al. (2010) Sandwich-cultured hepatocytes: an in vitro model to evaluate hepatobiliary transporter-based drug interactions and hepatotoxicity. Drug Metab. Rev. 42, 446–471 11 Hatherell, K. et al. (2011) Development of a three-dimensional, all-human in vitro model of the blood–brain barrier using mono-, co-, and tri-cultivation Transwell models. J. Neurosci. Methods 199, 223–229 12 Hermanns, M.I. et al. (2004) Lung epithelial cell lines in coculture with human pulmonary microvascular endothelial cells: development of an alveolo-capillary barrier in vitro. Lab. Invest. 84, 736–752 13 Sceats, E. (2010). In vitro tissue models: working in the third dimension. Innovat. Pharm. Technol. 28–32 14 Elliott, N.T. and Yuan, F. (2011) A review of three-dimensional in vitro tissue models for drug discovery and transport studies. J. Pharm. Sci. 100, 59–74 www.drugdiscoverytoday.com 933 REVIEWS Reviews  FOUNDATION REVIEW 15 Pörtner, R. et al. (2005) Bioreactor design for tissue engineering. J. Biosci. Bioeng. 100, 235–245 16 Radisic., M. et al. (2005) Mathematical model of oxygen distribution in engineered cardiac tissue with parallel channel array perfused with culture medium containing oxygen carriers. Am. J. Physiol. Heart Circ. Physiol. 288, H1278–H1289 17 Wendt, D. et al. (2012) Bioreactors in tissue engineering: scientific challenges and clinical perspectives. Adv. Biochem. Eng. Biotechnol. 112, 1–27 18 Bilello, J.A. et al. (1994) Effect of 20 ,30 -didehydro-30 -deoxythymidine in an in vitro hollow-fiber pharmacodynamic model system correlates with results of doseranging clinical studies. Antimicrob. Agents Chemother. 38, 1386–1391 19 Usuludin, S.B.M. et al. (2012) Co-culture of stromal and erythroleukemia cells in a perfused hollow fiber bioreactor system as an in vitro bone marrow model for myeloid leukemia. Biotechnol. Bioeng. 109, 1248–1258 20 Schmelzer, E. et al. (2009) Effect of human patient plasma ex vivo treatment on gene expression and progenitor cell activation of primary human liver cells in multicompartment 3D perfusion bioreactors for extra-corporeal liver support. Biotechnol. Bioeng. 103, 817–827 21 Zeilinger, K. et al. (2011) Scaling down of a clinical three-dimensional perfusion multicompartment hollow fiber liver bioreactor developed for extracorporeal liver support to an analytical scale device useful for hepatic pharmacological in vitro studies. Tissue Eng. C 17, 549–556 22 Park, J-K. and Lee, D-H. (2005) Bioartificial liver systems: current status and future perspective. J. Biosci. Bioeng. 99, 311–319 23 Allen, J.W. et al. (2001) Advances in bioartificial liver devices. Hepatology 34, 447– 455 24 Matsumura, K.N. et al. (1987) Hybrid bioartificial liver in hepatic failure: preliminary clinical report. Surgery 101, 99–103 25 De Bartolo, L. et al. (2000) A novel full-scale flat membrane bioreactor utilizing porcine hepatocytes: cell viability and tissue-specific functions. Biotechnol. Prog. 16, 102–108 26 Naruse, K. et al. (1996) Development of a new bioartificial liver module filled with porcine hepatocytes immobilized on non-woven fabric. Int. J. Artif. Organs 19, 347– 352 27 Sussman, N.L. et al. (1994) Extracorporeal liver support. Application to fulminant hepatic failure. J. Clin. Gastroenterol. 18, 320–324 28 Ni, M. et al. (2011) Characterization of membrane materials and membrane coatings for bioreactor units of bioartificial kidneys. Biomaterials 32, 1465–1476 29 Gómez-Lechón, M.J. et al. (2007) Hepatocytes: the choice to investigate drug metabolism and toxicity in man: in vitro variability as a reflection of in vivo. Chem. Biol. Interact. 168, 30–50 30 Gómez-Lechón, M.J. et al. (2003) Human hepatocytes as a tool for studying toxicity and drug metabolism. Curr. Drug Metab. 4, 292–312 31 Baudoin, R. et al. (2011) Behavior of HepG2/C3A cell cultures in a microfluidic bioreactor. Biochem. Eng. J. 53, 172–181 32 Snouber, L.C. et al. (2013) Metabolomics-on-a-chip of hepatotoxicity induced by anticancer drug flutamide and its active metabolite hydroxyflutamide using HepG2/C3a microfluidic biochips. Toxicol. Sci. 132, 8–20 33 Wilkening, S. et al. (2003) Comparison of primary human hepatocytes and hepatoma cell line Hepg2 with regard to their biotransformation properties. Drug Metab. Dispos. 31, 1035–1042 34 Darnell, M. et al. (2012) In vitro evaluation of major in vivo drug metabolic pathways using primary human hepatocytes and HepaRG cells in suspension and a dynamic three-dimensional bioreactor system. J. Pharmacol. Exp. Ther. 343, 134–144 35 Ulvestad, M. et al. (2012) Evaluation of organic anion-transporting polypeptide 1B1 and CYP3A4 activities in primary human hepatocytes and HepaRG cells cultured in a dynamic three-dimensional bioreactor system. J. Pharmacol. Exp. Ther. 343, 145– 156 36 Darnell, M. et al. (2011) Cytochrome P450-dependent metabolism in HepaRG cells cultured in a dynamic three-dimensional bioreactor. Drug Metab. Dispos. 39, 1131– 1138 37 Allen, J.W. and Bhatia, S.N. (2002) Improving the next generation of bioartificial liver devices. Semin. Cell. Dev. Biol. 13, 447–454 38 Patzer, J.F., II et al. (1999) Novel bioartificial liver support system: preclinical evaluation. Ann. N. Y. Acad. Sci. 875, 340–352 39 Gregory, P.G. et al. (2000) In vitro characterization of porcine hepatocyte function. Cell Transplant. 9, 1–10 40 Donato, M.T. et al. (1999) Characterization of drug metabolizing activities in pig hepatocytes for use in bioartificial liver devices: comparison with other hepatic cellular models. J. Hepatol. 31, 542–549 41 Naik., S. et al. (1996) Isolation and culture of porcine hepatocytes for artificial liver support. Cell Transplant. 5, 107–115 42 Vinken., M. et al. (2012) Primary hepatocyte cultures as in vitro tools for toxicity testing: quo vadis? Toxicol. In Vitro 26, 541–544 934 www.drugdiscoverytoday.com Drug Discovery Today  Volume 18, Numbers 19/20  October 2013 43 Bader, A. et al. (1996) Tacrolimus (FK 506) biotransformation in primary rat hepatocytes depends on extracellular matrix geometry. Naunyn Schmiedebergs Arch. Pharmacol. 353, 461–473 44 Selden, C. and Hodgson, H. (2004) Cellular therapies for liver replacement. Transpl. Immunol. 12, 273–288 45 Block, G.D. et al. (1996) Population expansion, clonal growth, and specific differentiation patterns in primary cultures of hepatocytes induced by HGF/SF, EGF and TGF alpha in a chemically defined (HGM) medium. J. Cell Biol. 132, 1133–1149 46 Jasmund, I. et al. (2002) Cultivation of primary porcine hepatocytes in an OXY-HFB for use as a bioartificial liver device. Biotechnol. Progr. 18, 839–846 47 Van de Kerkhove, M.P. et al. (2004) Clinical application of bioartificial liver support systems. Ann. Surg. 240, 216–230 48 Ding, Y-T. and Shi, X-L. (2011) Bioartificial liver devices: perspectives on the state of the art. Front. Med. 5, 15–19 49 Mazariegos, G.V. et al. (2002) First clinical use of a novel bioartificial liver support system (BLSS). Am. J. Transplant. 2, 260–266 50 Sauer, I.M. et al. (2002) Concept for modular extracorporeal liver support for the treatment of acute hepatic failure. Metab. Brain Dis. 17, 477–484 51 Ulvestad, M. et al. (2011) OATP1B1/1B3 activity in plated primary human hepatocytes over time in culture. Biochem. Pharmacol. 82, 1219–1226 52 Elsby, R. et al. (2012) Understanding the critical disposition pathways of statins to assess drug–drug interaction risk during drug development: it’s not just about OATP1B1. Clin. Pharmacol. Ther. 92, 584–598 53 Lau, Y.Y. et al. (2007) effect of OATP1B transporter inhibition on the pharmacokinetics of atorvastatin in healthy volunteers. Clin. Pharmacol. Ther. 81, 194–204 54 Abbott, N.J. et al. (2006) Astrocyte–endothelial interactions at the blood–brain barrier. Nat. Rev. Neurosci. 7, 41–53 55 Huber, J.D. et al. (2001) Molecular physiology and pathophysiology of tight junctions in the blood–brain barrier. Trends Neurosci. 24, 719–725 56 Cucullo, L. et al. (2005) Drug delivery and in vitro models of the blood-brain barrier. Curr. Opin. Drug Discov. Dev. 8, 89–99 57 Hori, S. et al. (2004) Functional expression of rat ABCG2 on the luminal side of brain capillaries and its enhancement by astrocyte-derived soluble factor(s). J. Neurochem. 90, 526–536 58 Cucullo, L. et al. (2002) A new dynamic in vitro model for the multidimensional study of astrocyte–endothelial cell interactions at the blood–brain barrier. Brain Res. 951, 243–254 59 Stanness, K.A. et al. (1997) Morphological and functional characterization of an in vitro blood–brain barrier model. Brain Res. 771, 329–342 60 Neuhaus, W. et al. (2006) A novel flow based hollow-fiber blood-brain barrier in vitro model with immortalised cell line PBMEC/C1-2. J. Biotechnol. 125, 127–141 61 Lacombe, O. et al. (2011) In vitro primary human and animal cell-based blood–brain barrier models as a screening tool in drug discovery. Mol. Pharm. 8, 651–663 62 Taylor, L. (2012) England’s ‘£1.4B price tag for kidney disease’. Pharma Times Online 9 August 63 Humes, H.D. and Szczypka, M.S. (2004) Advances in cell therapy for renal failure. Transpl. Immunol. 12, 219–227 64 Humes, H.D. et al. (1999) Tissue engineering of a bioartificial renal tubule assist device: In vitro transport and metabolic characteristics. Kidney Int. 55, 2502–2514 65 Saito, A. et al. (2006) Present status and perspective of the development of a bioartificial kidney for chronic renal failure patients. Ther. Apher. Dial. 10, 342–347 66 Humes, H.D. et al. (2004) Initial clinical results of the bioartificial kidney containing human cells in ICU patients with acute renal failure. Kidney Int. 66, 1578–1588 67 Oo, Z.Y. et al. (2011) The performance of primary human renal cells in hollow fiber bioreactors for bioartificial kidneys. Biomaterials 32, 8806–8815 68 Chertow, G.M. and Waikar, S.S. (2008) Toward the promise of renal replacement therapy. J. Am. Soc. Nephrol. 19, 839–840 69 Zhang, H. et al. (2009) The impact of extracellular matrix coatings on the performance of human renal cells applied in bioartificial kidneys. Biomaterials 30, 2899–2911 70 Masereeuw, R. and Russel, F.G.M. (2010) Therapeutic implications of renal anionic drug transporters. Pharmacol. Ther. 126, 200–216 71 Lash, L.H. et al. (2006) Membrane transport function in primary cultures of human proximal tubular cells. Toxicology 228, 200–218 72 Verhulst, A. et al. (2008) Human proximal tubular epithelium actively secretes but does not retain rosuvastatin. Mol. Pharmacol. 74, 1084–1091 73 Pietig, G. et al. (2001) Properties and regulation of organic cation transport in freshly isolated human proximal tubules. J. Biol. Chem. 276, 33741–33746 74 Zhang, L. et al. (1998) Role of organic cation transporters in drug absorption and elimination. Ann. Rev. Pharmacol. Toxicol. 38, 431–460 75 Leslie, E.M. et al. (2005) Multidrug resistance proteins: role of P-glycoprotein, MRP1, MRP2, and BCRP (ABCG2) in tissue defense. Toxicol. Appl. Pharmacol. 204, 216–237 76 Enomoto, A. et al. (2002) Interaction of human organic anion transporters 2 and 4 with organic anion transport inhibitors. J. Pharmacol. Exp. Ther. 301, 797–802 77 Lash, L.H. et al. (2008) Drug metabolism enzyme expression and activity in primary cultures of human proximal tubular cells. Toxicology 244, 56–65 78 Racusen, L.C. et al. (1997) Cell lines with extended in vitro growth potential from human renal proximal tubule: characterization, response to inducers, and comparison with established cell lines. J. Lab. Clin. Med. 129, 318–329 79 Szakács, G. et al. (2008) The role of ABC transporters in drug absorption, distribution, metabolism, excretion and toxicity (ADME–Tox). Drug Discov. Today 13, 379–393 80 Humes, H.D. et al. (1997) The bioartificial renal tubule assist device to enhance CRRT in acute renal failure. Am. J. Kidney Dis. 30, S28–S31 81 Dolt, K.S. (2009) 14-P007 differentiation of human embryonic stem cells towards renal progenitors. Mech. Dev. 126, 2 REVIEWS 82 Batchelder, C.A. et al. (2009) Renal ontogeny in the rhesus monkey (Macaca mulatta) and directed differentiation of human embryonic stem cells towards kidney precursors. Differentiation 78, 45–56 83 Humes, H.D. et al. (2002) Metabolic replacement of kidney function in uremic animals with a bioartificial kidney containing human cells. Am. J. Kidney Dis. 39, 1078–1087 84 Sato, Y. et al. (2005) Evaluation of proliferation and functional differentiation of LLC-PK1 cells on porous polymer membranes for the development of a bioartificial renal tubule device. Tissue Eng. 11, 1506–1515 85 Nedredal, G.I. et al. (2007) Significant contribution of liver nonparenchymal cells to metabolism of ammonia and lactate and cocultivation augments the functions of a bioartificial liver. Am. J. Physiol. 293, G75–G83 86 Duan, Y. et al. (2008) Shear-induced reorganization of renal proximal tubule cell actin cytoskeleton and apical junctional complexes. Proc. Natl. Acad. Sci. U.S.A. 105, 11418–11423 www.drugdiscoverytoday.com 935 Reviews  FOUNDATION REVIEW Drug Discovery Today  Volume 18, Numbers 19/20  October 2013